A Survey on Deep Learning Methods for Semantic Image Segmentation in Real-Time

Semantic image segmentation is one of fastest growing areas in computer\nvision with a variety of applications. In many areas, such as robotics and\nautonomous vehicles, semantic image segmentation is crucial, since it provides\nthe necessary context for actions to be taken based on a scene understanding at\nthe pixel level. Moreover, the success of medical diagnosis and treatment\nrelies on the extremely accurate understanding of the data under consideration\nand semantic image segmentation is one of the important tools in many cases.\nRecent developments in deep learning have provided a host of tools to tackle\nthis problem efficiently and with increased accuracy. This work provides a\ncomprehensive analysis of state-of-the-art deep learning architectures in image\nsegmentation and, more importantly, an extensive list of techniques to achieve\nfast inference and computational efficiency. The origins of these techniques as\nwell as their strengths and trade-offs are discussed with an in-depth analysis\nof their impact in the area. The best-performing architectures are summarized\nwith a list of methods used to achieve these state-of-the-art results.\n